I'm removing old targets that no longer appear actively maintained, to reduce the burden for target-independent codegen maintenance. Does anyone object to the removal of the Sparc backend? Dan
On 10/24/2011 05:45 PM, Dan Gohman wrote:> I'm removing old targets that no longer appear actively maintained, > to reduce the burden for target-independent codegen maintenance. > > Does anyone object to the removal of the Sparc backend? > > Dan > > _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdevPeople have been working on the Sparc backend recently. Sometimes these older backends are also good examples of how to code something in a new port. I based the Mips delay slot optimization on that in the Sparc port. I don't know about the alpha. That machine is fairly dead as far as I know. Sparc is not dead. Don't delete that. Reed
> I'm removing old targets that no longer appear actively maintained, > to reduce the burden for target-independent codegen maintenance.I don't know what you exactly mean. Could you give an example?> Does anyone object to the removal of the Sparc backend?Writing an LLVM Compiler Backend http://llvm.org/docs/WritingAnLLVMBackend.html It use SPARC as an example. If SPARC is removed from LLVM, then the tutorial must be rewritten. Regards, chenwj -- Wei-Ren Chen (陳韋任) Computer Systems Lab, Institute of Information Science, Academia Sinica, Taiwan (R.O.C.) Tel:886-2-2788-3799 #1667
> I'm removing old targets that no longer appear actively maintained, > to reduce the burden for target-independent codegen maintenance.But these targets often provide very valuable "live" documentation! Especially since the LLVM documentation does not provide a complete overview of features at all. If you remove them from the master, they should be available for reference somewhere else. I particularly object to the removal of sparc, which uses a lot of features and is a very important backend reference. Regards, Johannes
Dan Gohman wrote:> I'm removing old targets that no longer appear actively maintained, > to reduce the burden for target-independent codegen maintenance. > > Does anyone object to the removal of the Sparc backend? >I have an interest in the SPARC backend (and contributed a patch to improve it a few months ago, although I don't think the patch ever got applied). Other projects have got in the way, but I still intend to come back to this. Martin
On 10/25/2011 03:45 AM, Dan Gohman wrote:> I'm removing old targets that no longer appear actively maintained, > to reduce the burden for target-independent codegen maintenance. > > Does anyone object to the removal of the Sparc backend?I was hoping that someone would eventually (re)add JIT support to the Sparc backend, and then I would definetely use it in ClamAV. If the Sparc backend gets removed I don't think anyone will ever bother to do that, and then LLVM will only have 2 backends useful to me (with a working JIT): X86 (32/64), and PowerPC (32/64). According to my estimates Sparc is more widely used for ClamAV than PowerPC, so having LLVM support it would be nice. Best regards, ---Edwin
> If the Sparc backend gets removed I don't think anyone will ever bother to do that, and then LLVM will > only have 2 backends useful to me (with a working JIT): X86 (32/64), and PowerPC (32/64).LLVM has a working PowerPC JIT? -- Wei-Ren Chen (陳韋任) Computer Systems Lab, Institute of Information Science, Academia Sinica, Taiwan (R.O.C.) Tel:886-2-2788-3799 #1667
Dan Gohman <gohman <at> apple.com> writes:> > I'm removing old targets that no longer appear actively maintained, > to reduce the burden for target-independent codegen maintenance. > > Does anyone object to the removal of the Sparc backend?Removing the SPARC backend would be a pity. We use it in both teaching and research on embedded systems, targeting the LEON3 processor from Aeroflex Gaisler [1]. In addition to the other objections raised in this thread, there are ongoing efforts to build the RTEMS operating system with clang [2]. Quoting [2]:> clang does not support most of the architectures > supported by RTEMS. The intersection is x86, > mips, powerpc, sparc, sparcv9, arm, thumb, and blackfin. > In some cases, it does not cover the architectural > variants we care most about. For example, the > space community uses hardened SPARC V7 CPUs a lot > with RTEMS and only V8 is supported by clang right now.Kind Regards, Benedikt Huber [1] http://www.gaisler.com/cms/index.php?option=com_content &task=section&id=4&Itemid=33 [2] http://comments.gmane.org/gmane.comp.compilers.clang.devel/15630
On Oct 24, 2011, at 7:54 PM, 陳韋任 wrote:>> I'm removing old targets that no longer appear actively maintained, >> to reduce the burden for target-independent codegen maintenance. > > I don't know what you exactly mean. Could you give an example?I am making some scheduler changes. It appears that some of my changes uncovered a long-standing bug in the Alpha backend. I'm now debugging the Alpha backend. This takes effort. Dan